1

Unprecedentedly efficient CUG initiation of an overlapping reading frame in POLG

mRNA yields novel protein POLGARF

G Loughran1†, AV Zhdanov1†, MS Mikhaylova2†, FN Rozov3, PN Datskevich3, SI Kovalchuk4, MV

Serebryakova2, S Kiniry1, AM Michel1, PBF O’Connor1, DB Papkovsky1, JF Atkins1, PV Baranov1,4*,

IN Shatsky2*, DE Andreev2,4*

1 School of Biochemistry and Cell Biology, University College Cork, Cork, Ireland

2 Belozersky Institute of Physico-Chemical Biology, Lomonosov Moscow State University, Moscow,

Russia

3 Biological faculty, Lomonosov Moscow State University, Moscow, Russia

4 Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry, Moscow, Russia

* To whom correspondence should be addressed. Tel: +7 (926) 3987206; Fax: +7 (095) 9390338;

Email: cycloheximide@yandex.ru (DEA). Correspondence may also be addressed to

ivanshatskyster@gmail.com (INS), p.baranov@ucc.ie (PVB)

† Joint Authors

Classification

Biological Sciences/Genetics

Keywords

5’UTR, MIR, transposon, dual coding gene, POLG

Author Contributions

Conceptualization: DEA, Funding acquisition: INS, PVB, DEA, JFA. Investigation: GL, AVZ, MSM,

SK, SIK, AMM, PBFO, FNR, PND, MVS, PVB, DEA. Methodology: DEA, GL, AVZ, MSM, PVB,

INS, JFA, DBP. Project administration: DEA. Supervision: DEA, PVB, INS. Visualization: DEA, GL,

AVZ, MSM, MVS, PVB. Writing original draft: GL, AVZ, MSM, PVB, DEA, INS. Writing – review &

editing: all authors.

Abstract

While near cognate codons are frequently used for translation initiation in eukaryotes, their

efficiencies are usually low (<10% compared to an AUG in optimal context). Here we describe a

rare case of highly efficient near cognate initiation. A CUG triplet located in the 5’ leader of POLG

mRNA initiates almost as efficiently (~60-70%) as an AUG in optimal context. This CUG directs

translation of a conserved 260 triplet-long overlapping ORF, which we call POLGARF (POLG

Alternative Reading Frame). Translation of a short upstream ORF 5’ of this CUG governs the ratio

between DNA polymerase and POLGARF produced from a single POLG mRNA. Functional

investigation of POLGARF points to extracellular signalling. While unprocessed POLGARF resides

2

in the nucleoli together with its interacting partner C1QBP, serum stimulation results in rapid

secretion of POLGARF C-terminal fragment. Phylogenetic analysis shows that POLGARF evolved

~160 million years ago due to an MIR transposition into the 5’ leader sequence of the mammalian

POLG gene which became fixed in placental mammals. The discovery of POLGARF unveils a

previously undescribed mechanism of de novo protein-coding gene evolution.

Significance Statement

In this study, we describe previously unknown mechanism of de novo protein-coding gene

evolution. We show that the POLG gene, which encodes the catalytic subunit of mitochondrial DNA

polymerase, is in fact a dual coding gene. Ribosome profiling, phylogenetic conservation, and

reporter construct analyses all demonstrate that POLG mRNA possesses a conserved CUG codon

which serves as a start of translation for an exceptionally long overlapping open reading frame (260

codons in human) present in all placental mammals. We called the protein encoded in this

alternative reading frame POLGARF. We provide evidence that the evolution of POLGARF was

incepted upon insertion of an MIR transposable element of the SINE family.

Main Text

Introduction

The process of translation can be described in four steps: initiation; elongation; termination

and ribosome recycling. It is believed that protein synthesis is mostly regulated at the level of

initiation. In eukaryotes, the scanning model for translation initiation postulates that the small

ribosomal subunit, in complex with initiation factors and Met-tRNAi, enters at the 5’ end of mRNA

and then scans towards the 3’ end (1). Base-pairing interactions between the anticodon of the Met-

tRNAi and an AUG codon in the mRNA halts ribosome scanning and sets the reading frame for

subsequent elongation steps (2). Notably, due to mRNA mispairing with the anticodon of Met-

tRNAi, initiation can also occur at most triplets that differ from AUG by a single nucleotide (near-

cognate), albeit with much lower efficiency (3).

Initiation efficiency on any translation initiation site (TIS) critically depends on its

surrounding nucleotide context. In pioneering work, Kozak proposed that the context comprising

6 nt before and 1 nt immediately following a potential initiation codon has significant influence on

the recognition of an initiation site (4). In agreement with Kozak, recent high-throughput analysis of

all possible initiation contexts revealed RYMRMVAUGGC as the optimal context in human and

mouse cells and additionally revealed synergistic effects of neighboring nucleotides (5).

3

TISs in unfavorable context can be bypassed by the scanning ribosome in a process known

as leaky scanning. Since many mammalian mRNA 5’ leaders possess AUG codons, as well as

many potential near-cognate start codons, then leaky scanning must be widespread. However, the

mere presence of a potential TIS in a 5’ leader doesn’t necessarily guarantee initiation there. Until

recently, it was difficult to estimate how frequently upstream TISs (uTISs) are recognized by

scanning ribosomes in living cells. This can now be directly addressed since the emergence of the

ribosome profiling technique (riboseq), which allows monitoring of global translation at single

nucleotide resolution (6). Riboseq revealed widespread translation in the 5’ leaders of mRNAs,

especially in mammalian cells (7-9).

What is the role of translation initiation in 5’ leaders? In some instances it gives rise to N-

terminal extensions (10-12) though in most cases they result in translation of short (some are simply

AUG-stop) upstream open reading frames (uORFs). While it is believed that most uORFs suppress

translation of their main protein-coding sequence (13-15), a number of uORFs are involved in more

specialized regulation of translation, ranging from selective stress responses to eIF2

phosphorylation (16-19), to metabolite sensing (20-22).

Our attention was drawn to one such uORF within a mRNA encoding a catalytic subunit of

mammalian mitochondrial DNA polymerase (POLG). POLG is a hotspot for more than 200 known

mutations in humans that cause mitochondria-associated diseases such as PEOA1, SANDO, AHS,

and MNGIE (https://tools.niehs.nih.gov/polg/). Disease development is believed to result from a

gradual depletion of mtDNA due to polymerase dysfunction(s). Transgenic mice with a mutated

POLG causing proofreading deficiency (Polg mutator mouse) develop a mtDNA mutator

phenotype, which is characterized by low mtDNA copy number, decreased lifespan and premature

ageing (23).

There is a single AUG within the POLG 5’ leader which is expected to initiate translation of

a conserved 23 codon uORF. Here we show that, contrary to expectations, the removal of the

upstream AUG suppresses translation of the POLG coding sequence. Exploring the unusual effect

of the uORF mutation revealed highly efficient CUG initiation of a 260-codon long alternative

reading frame (-1) overlapping the POLG main ORF. Thus, the POLG mRNA turns out to be a dual

coding messenger.

Results

A CUG codon located upstream of POLG CDS governs translation of a long overlapping

reading frame that encodes POLGARF

The 5’ leader of the POLG mRNA contains a 23-codon conserved AUG-initiated upstream

uORF. To determine whether translation of this uORF affects the synthesis of POLG we fused the

4

whole 5’ leader of POLG (plus 33 nt downstream of the main ORF start codon) to a Firefly luciferase

(Fluc) reporter and explored the effect of preventing uORF translation on reporter activity (Fig. 1A).

In general, uORF translation represses translation of the main ORF by decreasing the number of

scanning 43S preinitiation complexes that reach the main ORF start codon (15), however, here,

translation of the uORF enhances main ORF translation (Fig. 1A). In search for potential

explanations we examined publicly available ribosome profiling data (24). Within the POLG mRNA,

the phase of triplet periodicity of ribosome footprints supports translation of an alternative reading

frame (frame -1) that overlaps the POLG CDS (Fig. 1B). This footprint density is higher than the

density of footprints aligning to the POLG CDS and decreases abruptly at the first stop codon in

the -1 frame located in exon 3 of POLG (Fig. 1C). Notably, this -1 frame stop codon is universally

conserved across placental mammals (Fig. 1C, middle panel) and the pattern of synonymous

substitutions in the POLG reading frame before the -1 frame stop codon suggests dual coding in

this region.

Since there are no AUG triplets that could initiate translation of the -1 frame ORF we

searched for conserved near-cognate initiation codons and, in exon 2, identified a CUG triplet

located 52 nt upstream of the POLG CDS start codon (Fig. 2A). To test whether this CUG codon

can initiate translation in the alternative reading frame, we fused the 5’ leader of POLG to Fluc in

the -1 frame (Fig. 2B). We observed robust -1 frame translation that increased by ~30% when

translation of the uORF was abolished (Fig. 2B, constructs 1 and 2). Thus, translation of the uORF

decreases -1 frame translation and increases 0 frame (POLG) translation (Fig. 1A), which is

consistent with preferential translation reinitiation in the 0 frame after uORF translation (Fig. 1A and

Fig. 2B). Replacement of the predicted CUG start codon with a non-initiating CUA completely

abolished -1 frame translation, strongly suggesting that this CUG triplet is the only -1 frame initiation

codon (Fig. 2B, construct 3). We termed this long ORF (260 codons in humans), which extensively

overlaps with the annotated POLG reading frame and starts at CUG, as POLGARF (POLG

Alternative Reading Frame).

The POLGARF CUG acts as a highly efficient initiation codon

Notably, in reporter constructs, initiation at the POLGARF CUG is ~60% as efficient as an

AUG in the same position (Fig 2B, constructs 1 and 4), which is markedly higher than most of the

values reported for CUG initiation (5-10% efficiency, constructs 6 and 8) (25). The exceptionally

high POLGARF CUG initiation efficiency was further explored by testing a series of 5’- and 3’-

truncated constructs. We found that almost the entire 5’ leader is dispensable for efficient -1 frame

initiation (Fig. S1). It has been reported that initiation at near cognate codons may be more

dependent on local context than initiation at AUG codons (26). Therefore, it seems highly probable

5

that such efficient POLGARF initiation is heavily reliant on its surrounding nucleotide context.

Particularly important is a G at the +4 position (where the first nucleotide of the start codon is +1)

which has been recently shown to be critical for near-cognate initiation efficiency (26). In the context

of the POLGARF CUG, the -3A and +4G are evolutionary conserved (Fig. 2A). To check whether

this purine nucleotide signature makes the observed CUG initiation so remarkably efficient, we

exchanged -3A to G and +4G to A in CUG- and AUG-initiating reporters. This exchange reduced

translation initiated at the AUG codon by 30% (Fig. S2), but CUG-initiated translation dropped by

85%, almost down to the levels expected for a “standard” CUG driven reporter (Fig. 2C).

Since all near-cognate initiation codons are expected to be suboptimal, it seemed likely

that the POLGARF CUG initiation should be sensitive to the levels of initiation factors responsible

for the stringency of sub-optimal start codon selection, i.e. EIF1 and EIF5 (27, 28). To test this, we

co-expressed POLG 5’ leader reporters with an excess of either EIF1 or EIF5. Surprisingly, we saw

almost no effect of EIF1 or EIF5 overexpression on POLGARF CUG-initiation: Thus in its natural

nucleotide context the POLGARF CUG codon behaves as a canonical AUG codon in optimal Kozak

context which is also unresponsive to EIF1 and EIF5 overexpression (29). However, the exchange

of -3G and +4A completely abolishes its independence from EIF1 and EIF5 levels. (Fig. 2C). Thus,

the nucleotide context of the CUG start codon rather than its identity is the major determinant of

POLGARF initiation efficiency. This context is critically dependent on the combination of -3A and

+4G and is also regulated by more distal nucleotides (Fig. S3).

What is the role of the short uORF upstream of POLGARF CUG? It is likely that ribosomes

access the POLGARF CUG either by leaky scanning past the uORF AUG and/or by reinitiation of

ribosomes that have translated the uORF. To determine whether leaky scanning is important for

POLGARF CUG-initiation we tested reporters in which reinitiation on CUG is prevented by

extending the uORF from 23 to 71 codons. Compared to wild-type constructs, CUG initiation is

reduced by >50% suggesting that leaky scanning plays an important role in POLGARF CUG

initiation (Fig. S4). In accordance with this, reducing leaky scanning by inserting a second in-frame

AUG within the extended uORF almost completely abolishes CUG initiation. Confirmation that

reinitiation plays an equally important role in CUG initiation is observed in constructs in which leaky

scanning is abolished by insertion of an in-frame AUG but reinitiation is still possible. Here we still

observe CUG initiation that is approximately half of wild-type CUG initiation. Furthermore, reducing

uORF length is thought to be more permissive for reinitiation and in agreement with this we observe

an approximately 10% increase in CUG initiation by decreasing the uORF from 23 to 5 codons. In

conclusion, both leaky scanning and reinitiation at the uORF appear to play an equally important

role in POLGARF CUG-initiation (Fig. S4).

6

Being originated through MIR transposition, POLGARF is conserved in placental mammals

To explore whether POLGARF encodes a functional protein we carried out phylogenetic

analysis of 100 vertebrate genomes (30). The POLGARF ORF as well as its CUG start codon are

conserved in placental mammals, and PhyloCSF analysis (31) reveals strong purifying selection

acting on the evolution of the POLGARF protein coding sequence (Fig. 3A).

RepeatMasker (http://repeatmasker.org) identified an MIR (mammalian-wide interspersed

repeat) that overlaps with the CUG codon and the short uORF (Fig. 3A). Detailed analysis of

vertebrate alignments suggests that POLGARF originated by an MIR transposition before adaptive

radiation of placental mammals (Fig. 3B and discussion below).

Detection of endogenous POLGARF protein

Alternative reading frame translation does not necessarily mean the existence of a stable

protein product. To search for endogenous POLGARF peptides, we applied the post acquisition

targeted search (PATS) technique (32) to a BioPlex interactome dataset containing AP-MS results

for ~6000 protein baits overexpressed in HEK293 cells (33). The PATS algorithm predicted

POLGARF peptides in >90 protein baits, among which TRIP13, CAMK2D, NPM2, HAVCR2,

CLEC3A and CHCHD10 pull down datasets were predicted to have 2 or more POLGARF originated

peptides (Table S1).

Direct .raw data analysis for the latter baits indeed identified four POLGARF tryptic

peptides; three of these are not found within nr protein database by BLAST (34) and are unique for

POLGARF. Since this is the first ever observation of POLGARF protein product, we confirmed the

fidelity of peptide identification by comparing MS2 spectra from BioPlex .raw files with our LC-

MS/MS data produced with overexpressed POLGARF (Fig. S5). The pattern of tryptic peptide

coverage and MS2 spectra were highly similar between the two datasets. Together these data

unambiguously identify endogenous POLGARF protein in HEK293 cells from the BioPlex data.

Notably, the abundance of POLGARF in cells seems to be very low, as we were unable to

detect POLGARF originated peptides in either in-house generated proteomics data for HEK293

cells without POLGARF overexpression nor in deep proteome datasets from (35) for HEK293,

HeLa, colon, liver, HCT116 and prostate cells, while in the same datasets several POLGARF

originated peptides were detected in MCF7, SHSY, A549 cells as well as in Jurkat cells in the

dataset from (36)) and in immune cells in the dataset from (37). All peptide identifications are

isolated and with very low intensities. Considering the high efficiency of POLGARF translation from

POLG mRNA observed with RiboSeq and reporter assays, it seems likely that endogenous

7

POLGARF is either compartmentalised within subcellular locations that are insoluble under regular

cell lysis protocols, or is secreted from the cell or else unstable.

POLGARF protein interacts with C1QBP and redirects it to nucleoli

In order to shed light on POLGARF function, we searched for its interacting partners using

a GST-POLGARF fusion protein overexpressed in Expi293F cells. Pull down assays detected a 32

kDa major protein identified as C1QBP, a.k.a. P32 (Fig. 4A). The P32 homotrimer adopts a

doughnut-shaped quaternary structure with asymmetric charge distribution on its surface. P32 is

involved in a wide range of intracellular and extracellular activities. However, directed by a

mitochondrial targeting sequence (MTS), it predominantly localizes in the mitochondrial matrix (38).

In mitochondria, P32 is thought to control the translation of mitochondrially encoded proteins, either

directly or by affecting mitochondrial ribosome biogenesis (39, 40). We confirmed the specificity of

the POLGARF/P32 interaction using SNAP-tag pull down assays: SNAP-POLGARF and SNAP-

P32 fusion proteins efficiently pulled down P32 and POLGARF, respectively (Fig. 4B). Next, we

found that SNAP-POLGARF, when overexpressed in HEK293T cells, accumulates in nucleoli,

where it colocalizes with fibrillarin, one of the major nucleolar components (Fig. 4C), and does not

colocalize with SC35, a nuclear speckle marker (Fig. S6). Nucleolar localization of POLGARF is

dependent on amino acids located within its N-terminal half (Fig. S7). In nucleoli, POLGARF

localizes to areas of active rRNA production, enriched with RNA polymerase I (Fig. 4C). SNAP-

P32 alone was not observed in the nucleoli, however, when co-expressed with CLIP-POLGARF,

SNAP-P32 showed clear nucleolar localization (Fig. 4D). To determine whether it is full-length, or

mature P32 (without MTS) that accumulates in the nucleoli in a POLGARF-dependent manner, we

carried out subcellular fractionation of cells overexpressing P32 with or without POLGARF (Fig.

4E). Mass spectrometry analysis of nucleolar P32 demonstrated that it retained the MTS (Fig. S8).

This suggests that its interaction with POLGARF prevents P32 maturation, redirects P32 from the

mitochondria to the nucleoli and thus may affect P32 functions.

POLGARF C-terminal fragment, POLGARFin, is secreted from cells upon serum stimulation

To investigate the kinetics of POLGARF accumulation, we fused POLGARF with HiBiT, an

11 amino acid peptide which can complement a truncated Nanoluciferase fragment (LgBiT) to

regain full activity (41). After transfection of POLGARF-HiBit we detected a progressive increase in

luciferase activity in Hek293T cell lysates. Interestingly, we also detected luciferase activity in the

conditioned media (Fig. S9). Extracellular HiBiT-containing protein was purified from the

conditioned media with an engineered SNAP-LgBiT protein and SNAP magnetic beads. Upon

fractionation by SDS-PAGE intracellular POLGARF-HiBiT migrates at 35 kDa as expected (Fig 5A),

whereas an extracellular HiBiT fusion protein migrates at ~17 kDa (Fig 5B). LC-MS analysis of the

8

extracellular HiBit fusion identified it as a heterogenenous population of C-terminal POLGARF

fragments mainly produced by cleavages around positions 138-140 and 150 with a minor

population sprinkle both longer and shorter proteoforms around the major cleavage positions (Fig.

5C). We called these fragments POLGARFin. Notably, POLGARFin fragments were not observed

in cell lysates representing soluble cytosolic fractions, nor could we detect full length POLGARF in

the media (Fig 5A). It seems unlikely that POLGARFin-HiBiT is released into the media from dead

cells. The most probable explanation is that POLGARFin is secreted immediately after intracellular

POLGARF cleavage. Alternatively, the full-length protein can be secreted and immediately cleaved

outside of cells.

To investigate whether this secretion is regulated, we subjected POLGARF-HiBiT

overexpressed cells to various treatments. It appeared that POLGARFin secretion is upregulated

by serum addition: when the media of transfected cells is supplemented with fresh media containing

10% FBS, rapid accumulation of POLGARFin is observed within 30 minutes after stimulation with

no further increase over time (Fig 5C). In contrast, addition of serum free media does not result in

POLGARFin secretion. POLGARFin secretion in response to FBS-containing media does not

depend on de novo protein synthesis, as supplementation of cells with serum-rich media infused

with cycloheximide does not prevent POLGARFin extracellular accumulation. Collectively, these

data suggest that POLGARF is processed into actively secreted POLGARFin, which may be

implicated in extracellular signalling events.

Discussion

The results presented above provide strong evidence for the unusual existence of a novel

functional protein encoded within the POLG mRNA. When two different proteins are translated from

a single mRNA, the efficiency of initiation at each start codon could set the ratio of product steady

state levels assuming each protein has similar stability. However, this ratio may vary during

conditions in which initiation efficiency is altered. Eukaryotes have developed elaborate

mechanisms for the recognition of the correct initiation codon and the levels of certain initiation

factors can regulate the fidelity of initiation, especially on sub-optimal (near-cognate and AUG in

poor context) start codons (42). While elevated levels of EIF1 can increase the stringency of start

codon selection, elevated levels of EIF5 have the opposite effect. Here we show that

overexpression of either EIF1 or EIF5 had minimal effect on CUG-initiated POLGARF (Fig. 2C)

suggesting that this highly efficient CUG initiation is refractory to normal stringency controls.

Our analysis of the role of the conserved uORF in POLGARF initiation reveals that both

leaky scanning and reinitiating ribosomes can equally start translation at the POLGARF CUG (Fig.

9

S4). However, since preventing translation of the uORF (therefore no reinitiation) results in

predominantly CUG initiation then it follows that under normal conditions most of the ribosomes

initiating the POLG ORF are reinitiating ribosomes. This raises the intriguing possibility that the

ratio of POLG to POLGARF may be regulated by stress conditions.

Although the contribution of SINE transposons to eukaryotic gene evolution, including their

exonization (43), is well known, to our knowledge this is the first example of a retrotransposon

creating a dual coding gene. How did POLGARF evolve and why did it subsequently become fixed

in placental mammals? All mammalian sequences with the exception of platypus share significant

similarity with MIR sequences as determined with Dfam search (44). Therefore, it is likely that MIR

insertion occurred after Monotremata diverged from the common ancestor of Marsupials and

Placentals. We found that POLG sequences of many vertebrate species (including platypus) lack

stop codons in one of the alternative frames in the first coding exon, thus, acquisition of an in-frame

start codon could lead to the expression of the alternative frames. However, MIR did not contain a

suitable start codon. Two subsequent mutations had to occur to enable POLGARF expression: a

substitution of CCA with CTG and a single nucleotide deletion downstream that was necessary to

place the CTG in-frame with the alternative ORF. Interestingly, these variants are common for both

marsupials and placental mammals (shown in green in Fig 3B), suggesting that proto-POLGARF

existed in their common ancestor. However, marsupials have two variants (in red in the region

shown in Fig 3B) and several stop codons in the POLGARF frame further downstream, suggesting

that POLGARF was subsequently lost in marsupials, while in the common ancestor of placental

mammals it acquired a functional role and became fixed in subsequent lineages.

What may be the functional role of POLGARF? The polypeptide can be tentatively divided

into four parts; notably, the 64 amino acid long C-terminus is the most conserved region with 22

invariant amino acids (Fig. S10). We failed to find any significant similarity between POLGARF and

other known or predicted proteins or any similarity with known structural motifs. It seems likely that

POLGARF is an intrinsically disordered protein (IDP) with a remarkably high isoelectric point (pI

=12.05 for a human protein).

As a conserved IDP, POLGARF has a good potential to be an important regulatory protein.

In cell signalling and regulation, IDPs emerged as parts of integrated circuits. Indeed, the capacity

of IDPs to acquire numerous posttranslational modifications and change conformation in a context-

dependent manner allows immense versatility of their interactomes (45, 46). The observed specific

interaction of POLGARF with P32 as well as the modulatory effects of POLGARF on P32

localisation and processing exemplify the importance of POLGARF for cell functioning (Fig 4 and

10

Fig S8). Along with P32, other putative interaction partners of POLGARF (TRIP13, CAMK2D, etc.,

see Table S1) can be considered as good candidates for mechanistic follow up studies.

According to RiboSeq analysis and reporter assays, the levels of POLG and POLGARF

proteins should be comparable. However, according to proteomics data, in contrast to the

moderately high levels of the housekeeping protein POLG, POLGARF’s concentration is extremely

low. Our discovery that POLGARFin can be secreted is likely the most probable explanation for

this discrepancy. We propose that endogenous POLGARF may be almost completely processed

and secreted outside of cells, where it may participate in currently unknown cell-to-cell

communication. We speculate that the levels of secreted POLGARFin reflects the capacity of the

donor cell to have enough POLG to replicate mtDNA, as both POLG and POLGARF are encoded

in the same mRNA. In contrast, cells with decreased expression of POLG mRNA would also

produce less POLGARFin.

Finally, these findings could have profound implications for the interpretation of POLG

mutations. As a previously unknown dual coding gene, POLG could bear hidden mutations

responsible for diseases of still unknown aetiology. Among the known POLG mutations some do

not cause changes in the amino acid composition of POLG. Such synonymous single nucleotide

variations (SNVs) are often considered harmless (except for their potential effects on splicing) (47).

However, in dual coding regions (such as POLG/POLGARF described in this study), synonymous

SNVs in one ORF are unlikely to be synonymous in the other ORF. This emphasizes the need to

consider POLGARF variants in future studies.

Materials and Methods

Cell culture

Here and elsewhere, all reagents were from Millipore-Sigma, unless stated otherwise.

Human embryonic kidney HEK293T and Expi293F cells were from the American Collections of Cell

Cultures (Manassas, VA) and Thermo Fisher Scientific (Rockford, Ill), respectively. HEK293T were

maintained in 75 cm2 flasks (Sarstedt, Wexford, Ireland) in DMEM supplemented with 10% FBS,

2 mM L-glutamine, 100 U/ml penicillin / 100 g/ml streptomycin (complete DMEM) with or without

10 mM HEPES (pH 7.2), in humidified atmosphere of 95% air and 5% CO2 at 37C. Expi293F cells

were maintained in Expi293 Expression medium (Thermo Fisher) at 8% CO2 and 37C under

continuous mixing at 160 RPM on an orbital shaker.

Plasmids and constructs

11

To create plasmids for luciferase reporter assays, the POLG 5’ leader plus 33 nt of POLG

CDS were amplified by PCR from HEK293T cDNA using sense primer 5’-

GACCCAAGCTTAAGTTGCGGCTGCCAGCG and antisense primer 5’-

TTATTGGATCCCCGGCCACCTTCCTCCAGAGC with flanking HindIII and BamHI restriction

sites. All other variants were generated by 2-step PCR using appropriately designed mutagenic or

nested deletion primers. All PCR amplicons were digested with HindIII and BamHI and ligated into

HindIII and BamHI digested dual luciferase vector p2-Luc (48) to make pSV40-firefly reporter

constructs indicated. All constructs were verified by sequencing. Plasmids used to overexpress

deregulated eIF1 (eIF1g*), and deregulated eIF5 (eIF5-AAA), have been described previously (29).

Invitrogen pcDNA3 plasmid (Thermo Fisher) was used for control transfections.

To create plasmids for full length POLGARF-FLAG overexpression, the region of the POLG

mRNA corresponding to the complete 5’ leader to the POLGARF stop codon was amplified by PCR

from a HEK293T cDNA library using a sense primer containing a 5’ HindIII site and an antisense

primer containing FLAG in frame to POLGARF followed by an XbaI site. This amplicon was cloned

between HindIII and XbaI in pcDNA3.1 (Thermo Fisher) to create pcDNA3.1 POLGARF-FLAG. In

parallel, a plasmid without the full 5’leader and with a CUG to AUG mutation was created using the

same reverse primer containing FLAG (pcDNA3.1 POLGARF AUG-leader).

The POLGARF and P32 coding sequences were first amplified and inserted into Invitrogen

pGEX6p1 vector (Thermo Fisher) between BamH1 and NotI sites (in frame to N-terminal GST tag)

to create pGEX6p1-POLGARF and pGEX6p1-P32 respectively. These constructs were used to

prepare the subsequent constructs used in this study. First, the GST-POLGARF region was

amplified from pGEX6p1-POLGARF using a forward primer which modifies the GST start codon to

an efficient initiator in mammalian cells. This fragment had primer-incorporated XbaI and NotI sites

and was inserted into a modified pcDNA3.4 vector (Thermo Fisher), which contained a custom

polylinker (with XbaI and NotI restriction sites), thus yielding pcDNA3.4-GST-POLGARF.

To create constructs with SNAP and CLIP tags, the pGEX6p1-POLGARF and pGEX6p1-

P32 were digested with BamH1 and NotI and cloned into pSNAPf and pCLIPf vectors from New

England Biolabs (NEB, Ipswich, MA) to create pSNAPf-POLGARF, pCLIPf-POLGARF, pSNAPf-

P32 and pCLIPf-P32, respectively. These constructs were further re-cloned into pcDNA3.4 (Spe1-

Not1 digestion) to create pcDNA3.4-SNAP-POLGARF, pcDNA3.4-CLIP-POLGARF, pcDNA3.4-

SNAP-P32 and pcDNA3.4-CLIP-P32, respectively.

To create the plasmid containing POLGARF with the C-terminal HiBiT tag, pcDNA3.4-

POLGARF was modified by amplification of POLGARF with a reverse primer containing the HiBiT

12

sequence and NotI. The resulting fragment was digested with HindIII and NotI, then again inserted

in pcDNA3.4 to yield pcDNA3.4-POLGARF-HiBiT.

To create the plasmid containing SNAP-LgBiT, the gene block containing LgBiT sequence

flanked by BamH1 and NotI was ordered from IDT. pcDNA3.4-SNAP-POLGARF plasmid was

treated with BamH1 and NotI, and POLGARF coding fragment was exchanged with LgBiT

sequence to yield pcDNA3.4. SNAP-LgBiT.

The P32 CDS with C-terminal FLAG was cloned in pcDNA3.4 to create pcDNA3.4-P32-

Flag. The construct pcDNA3.1 POLGARF AUG-leader was recloned into pcDNA6 vector to create

pcDNA6-POLGARF-Flag.

The pcDNA3.4-SNAP-POLGARF construct was used to generate the series of N-terminal

and C-terminal deletions: N1 to N5 corresponds to progressive 26 aa deletions from its N-

terminus, while C1 to C5 corresponds to progressive 26 aa deletions from its C-terminus (Fig

S7). The sequences were amplified with corresponding primers bearing BamH1 and Not1

restriction sites and were cloned back into pcDNA3.4 to generate truncated fusion proteins.

Cell Transfection

Four main transfection protocols were used in different layouts.

1. One-day Lipofectamine 2000 based protocol, in which suspended cells are added directly

to the DNA complexes prepared in OptiMEM.

For luciferase assays, HEK293T cells were transfected in half-area 96-well or 6-well plates

(Sarstedt). For Figs. 1A, 2B and Supplementary Figs. 1-3, the following were added to each well:

25 ng of each pSV40-firefly vector, 5 ng control pSV40-Renilla vector plus 0.2 μl Lipofectamine

2000 in 25 μl OptiMEM (both from Thermo Fisher). For Fig. 2C, the following were added to each

well: 50 ng of protein overexpressing vector, 5 ng pSV40-firefly vector (with initiation contexts

and/or codons as indicated in the figures), 1 ng pSV40-Renilla vector and 0.2 μl Lipofectamine

2000 in 25 μl OptiMEM. The transfecting DNA complexes in each well were incubated with 3×104

cells suspended in 50 μl DMEM + 10% FBS at 37°C in 5% CO2 for 24 hr. For Western blotting

analysis of luciferase expression, 106 HEK293T cells suspended in 3 ml DMEM + 10% FBS were

transfected overnight with 1 g of each indicated plasmid in 6-well plates.

2. One-day Lipofectamine 2000 based protocol, in which DNA complexes prepared in

OptiMEM are added to 50-60% confluent adherent cells.

13

For confocal imaging and Western blotting analysis of antibodies against POLGARF,

HEK293T cells were seeded at 1.5×104 cells /cm2 on glass bottom dishes (MatTek, Ashland, MA))

or 2.5×105 per well of 6-well plates (Sarstedt). After 12 h incubation, plasmids encoding construct

as indicated in Figures were delivered in cells by 3-4 h incubation with DNA complexes; 100 ng

DNA and 0.4 l Lipofectamine were used per 1 cm2. Then transfection mixture was replaced with

standard medium and cells were grown for 20 h prior to lysis or SNAP/CLIP staining followed by

live cell microscopy or immunostaining.

3. FuGene protocol (40-48 h).

For GST- and SNAP-pull down assays, HEK293T cells were transfected with combinations

of pcDNA3.4-SNAP-POLGARF, pcDNA3.4-CLIP-POLGARF, pcDNA6-POLGARF-Flag,

pcDNA3.4-P32-Flag, and pcDNA3.4 encoding fluorescent protein (Fast Fluorescent Timer) (49) as

a negative control (2 g of DNA per well of 6 well plate) with FuGene transfection reagent

(Promega, Madison, WI) according to manufacturer’s instructions.

4. Expifectamine 293 protocol (48 h).

For subcellular fractionation, 2 × 108 expi293F cells were transfected with DNA. For that,

80 μg DNA and 216 μl Expifectamine 293 (Thermo Fisher) were prepared in 4 ml OptiMEM, mixed

together, incubated for 20 minutes at room temperature and added to the cells. In 24 h, enhancers

1 and 2 (400 μl /4 ml) and gentamicin (up to 10 μg/ml) were added. Cells were harvested by

centrifugation (10 min, 3000 g) 48 h after transfection.

Dual Luciferase Assay and Western analysis

Firefly and Renilla luciferase activities were determined using the Dual Luciferase Stop &

Glo® System (Promega). Relative light units were measured on a Veritas Microplate Luminometer

with two injectors (Turner Biosystems, Sunnyvale, CA). Transfected cells were lysed in 12.6 μl of

1 × Passive Lysis Buffer (PLB, Promega) and light emission was measured following injection of

25 μl of either Renilla or Firefly luciferase substrate. Initiation efficiencies (% initiation) were

determined by calculating relative luciferase activities (Firefly/Renilla) of test constructs and dividing

by relative luciferase activities from replicate wells of control ATG constructs.

For Western blotting analysis of luciferases, transfected cells were lysed in 100 l 1 × PLB.

Proteins were resolved by SDS-PAGE and transferred to Protran nitrocellulose membranes (GE

Healthcare Life Sciences, Waukesha, WI), which were incubated at 4°C overnight with primary

antibodies. Immunoreactive bands were detected on membranes after incubation with appropriate

14

fluorescently labelled secondary antibodies using a LI-COR Odyssey® Infrared Imaging Scanner

(LI-COR, Lincoln, NE).

For Western blotting analyses of POLGARF expression, transfected cells were washed

with PBS and lysed for 20 min on ice with RIPA buffer (Thermo Fisher), containing phosphatase

and protease inhibitors; complete protease inhibitor cocktail and PhosSTOP tablets were from

Roche (Mannheim, Germany). After lysate clarification by centrifugation for 15 min at 14000 g and

4°C, protein concentration was measured using BCATM Protein Assay kit (Thermo Fisher) and

equalised. Proteins were separated by 4-20% polyacrylamide gel electrophoresis using pre-made

acrylamide gels and running buffers from GeneScript (Piscataway, NJ), transferred onto a 0.2 μm

ImmobilonTM-P PVDF membrane (Sigma) using Hoefer™ TE 22 transfer system (Hoefer,

Holliston, MA) and probed with antibodies against POLGARF (1:1000) and -tubulin (1:5,000) in

5% fat-free milk in TBST (0.8% Tween-20). Immunoblots were analysed using Amersham™ ECL™

Prime Kit from GE Healthcare Life Sciences (Waukesha, WI) and the LAS-3000 Imager (Fujifilm,

Japan). Quantitative image analysis was performed with ImageJ program using -tubulin signals

for normalisation.

GST and SNAP pull down assays

For GST pull down assay, 2 × 108 expi293F cells transfected with pcDNA3.4-GST-

POLGARF or with pcDNA3.4-GST were harvested by centrifugation and lysed in 2 ml of PLB. The

lysates were diluted with PBS to 10 ml and incubated with 200 l of GST-Sepharose (GE

Healthcare) for 1 hour on ice under agitation. GST resin was washed 3 times with 10 ml of PBS,

and POLGARF bound proteins were eluted by incubation with 1 l of Prescission Protease (GE

Healthcare) in PBS at 4°C overnight. The eluted proteins were resolved by SDS-PAGE and stained

with Coomassie, protein bands were excised and subjected to MALDI TOF MS/MS analysis.

For SNAP pull down assay, transfected cells were lysed in 200 l PLB supplemented with

1 mM DTT. The lysates were mixed with 30 l of SNAP-magnetic beads (NEB) and incubated for

1 h at 24°C on a thermomixer (900 rpm). After incubation, the beads were washed twice with 1 ml

of PBS supplemented with 1 mM DTT, and bound proteins were eluted by boiling with SDS gel

loading buffer. The samples were resolved on SDS page and immunoblotted with either anti-FLAG

or with custom made anti-POLGARF antibodies.

Staining of cells with probes and confocal microscopy

Transfected cells were stained with SNAP-Cell® TMR-Star or SNAP-Cell 647-SiR (NEB),

both diluted 1:500 with complete DMEM, for 30 min immediately prior to live cell imaging or

15

immunostaining. Live cell imaging was conducted on an Olympus FV1000 confocal laser scanning

microscope with controlled CO2, humidity and temperature. Alexa-555 and SNAP-Cell® TMR-Star

probes were excited at 543 nm (1-10% of laser power) with emission collected at 560-600 nm.

Alexa-647 and SNAP-Cell 647-SiR dyes were excited at 633 nm (2-10% of laser power); emission

was collected at 640-700 nm. Separate spectral signals were collected with 0.5 m steps (80-120

m aperture) in sequential laser mode with emission gates adjusted to avoid spectral overlap.

Fluorescence and differential interference contrast (DIC) images were collected using oil immersion

UPLSAPO 60x/1.35 Super Apochromat objective. Analysis was performed using FV1000 Viewer

software (Olympus) and Microsoft Excel.

Immunofluorescence analysis

Immunostaining was performed according to a standard procedure. Briefly, cells were

washed with Dulbecco’s PBS (supplemented with Ca2+ and Mg2+) fixed with 3.7%

paraformaldehyde in PBS, quenched with 50 mM NH4Cl, permeabilised with 0.25% TX100, blocked

with 5% BSA in TBST and incubated with primary and secondary antibodies for 1 h (as indicated

in Figures, Supplemental Table 3). Standard TBST was used for cell washes. Commercial primary

antibodies were diluted in blocking solution 1:500-1:800; POLGARF antibodies - 1:300; secondary

antibodies - 1:1000. When required, cell nuclei were stained with DAPI. Cells were left in PBS for

confocal imaging performed within 4 h as described above for live cells. DAPI was excited at 405

nm with emission collected at 420-460 nm.

Subcellular Fractionation

Nucleolar fraction was prepared from suspension expi293F cells according to the protocol

described by Lamond and co-authors (50) with modifications. 2 × 108 cells were harvested by

centrifugation 48 h after transfection. Cell pellets were re-suspended in 2 ml of hypotonic buffer A

(10 mM HEPES (pH 7.9), 10 mM KCl, 1.5 mM MgCl2, 0.5 mM DTT and Complete Protease Inhibitor

tablet), incubated on ice for 15 min, homogenized 25 times in a Dounce tissue homogenizer (using

a tight pestle ‘B’) and centrifuged at 1000 g for 5 min. Pellets were re-suspended in buffer A,

incubated on ice for 10 min, then homogenization followed by centrifugation was repeated in order

to obtain purer nuclear pellet. Then the pellet was re-suspended in 3 ml of S1 solution (0.25 M

Sucrose, 10 mM MgCl2 and Complete Protease Inhibitor tablet), layered over 3 ml of S2 solution

(0.35 M Sucrose, 0.5 mM MgCl2 and Complete Protease inhibitor tablet), and centrifuged at 1430

g for 5 min. Purified nuclear pellets re-suspended in 3 ml of S2 solution were sonicated 5 x 10

seconds using SONICS Vibra cell (Sonics & Materials inc, Newton, CT) (50% of max power).

Sonicated sample was layered over 3 ml of S3 solution (0.88 M Sucrose, 0.5 mM MgCl2 and

16

Complete Protease Inhibitor tablet) and centrifuged at 3000 g for 10 min. The resulting pellet

contained the nucleolar enriched fraction.

MALDI TOF analysis

Protein spots were excised from the gel and digested with trypsin. The spots (3-4 mm3)

were destained with 50 mM ammonium bicarbonate 40% acetonitrile then dehydrated with 100 μL

acetonitrile (ACN) and rehydrated with 5 μL of digestion solution containing 20 mM ammonium

bicarbonate and 15 ng/μL sequencing grade trypsin (Promega). Digestion was carried out at 37°C

for 6 h. Peptides were extracted with 10 μL of 0.5% trifluoroacetic acid (TFA) solution. 1 μL of

extract was mixed with 0.5 μL of 2,5-dihydroxybenzoic acid saturated solution in 20% ACN and

0.5% TFA on the stainless steel MALDI sample target plate. Mass spectra were recorded on an

UltrafleXtreme MALDI-TOF/TOF mass spectrometer (Bruker Daltonics, Germany) equipped with

an Nd laser (354 nm). The MH+ molecular ions were detected in reflector-mode in the mass range

600-6000 m/z. The accuracy of mass peak measurement was within 30 ppm. Peptide

fragmentation spectra were obtained in LIFT mode with the accuracy of daughter ions

measurement within 1 a.m.u. range. Mass-spectra were processed in FlexAnalysis 3.3 software

(Bruker Daltonics) and analysed manually. Protein identification was performed by Mascot software

(Matrix Science, London, UK) against UniProt reference proteome (UP000005640) database.

Partial oxidation of methionine and propionamidomethylation of cysteine residues was assumed

for peptide mass fingerprint searches. Protein scores greater than 70 were considered as

significant (p< 0.05).

Identification of POLGARF in proteomics data

Using the recently developed Post-Acquisition Targeted Searches technique (32), we

submitted the 8 longest predicted tryptic peptides from POLGARF for a targeted search in the pre-

analysed HEK293 interactome dataset (33) through http://www.pepchem.org/PATS. The search

yielded a total of 118 .raw file hits for 94 protein baits (Table S1). For the baits with >2 predicted

peptide identifications, all .raw data were downloaded from BioPlex2.0 website

(http://bioplex.hms.harvard.edu) for in-house database search analysis.

POLGARF predicted tryptic peptides were searched in PeptideAtlas (51). One of the

peptides AAAAQPIGHPDALER turned out to be known and identified in several proteome datasets

mostly connected with immune cells. In particular this peptide was identified in Jurkat dataset from

(36) and in immune cell analysis from (37). For some other cell lines, including HEK293 cells, the

dataset was taken from (Bekker-Jensen et al., 2017). Corresponding datasets were downloaded

from ProteomeExchange for in-house database search analysis. For the list of all publicly available

17

datasets downloaded and reanalysed against POLGARF-containing protein database see

Supplementary Table 4.

To obtain reference peptide spectra, POLGARF was overexpressed in HEK293T and HEK

293F cells. 293T cells were transfected in 12 well plates with 1 ug of pcDNA3.4 POLGARF or

pcDNA3.4 Timer (as a negative control) with Fugene transfection reagent (Promega) according to

manufacturer’s instructions. After 48hr the cells were washed with PBS and lysed in a sonication

bath in 150 ul of 1x Passive Lysis Buffer (Promega) supplemented with Protease Inhibitor cocktail

(Roche). Expi293F cells were transfected with pcDNA3.4 POLGARF or pcDNA3.4 (40 ug of DNA

per 108 of cells) with Expifectamine 293 (Thermo Fisher). After 48hr cells were washed with PBS

and lysed in a sonication bath in 1x Passive Lysis Buffer (Promega) supplemented with Protease

Inhibitor cocktail (Roche). The lysates were heated for 10 min at 90oC. Protein material was purified

by MeOH-chloroform precipitation (52) and resuspended in 1% SDC, 100 mM Tris-HCl, pH 8.5

buffer. Protein concentration was measured by microBCA assay and 50 g of protein was

readjusted to the concentration 1 mg/ml with the 1% SDC, 100 mM Tris-HCl, pH 8.5 buffer, cysteine

disulphide bonds were reduced and alkylated by adding TCEP and CAA to 10 and 20 mM

correspondingly and heating at 85oC for 10 min. For protein digestion, trypsin (Promega, USA) was

added at 1/100 w/w ratio twice with the first digestion for 2 h and the second digestion overnight at

37oC. Digestion was stopped by adding TFA to 1%. Precipitated SDC was removed by

centrifugation. The samples were directly analysed by LC-MS without SPE.

LC-MS analysis was done on an Ultimate 3000 RSLCnano HPLC system connected to a

QExactive Plus mass spectrometer (Thermo Fisher). Samples were loaded on a home-packed 2

cm x 100 m precolumn packed with Inertsil ODS3 3 m sorbent (GLSciences, Japan) in loading

buffer (2% ACN, 98% H2O, 0.1% TFA) at 10 l/min flow rate and separated on a 50 cm x 100 m

analytical column home-packed (53) with Reprosil Pur C18 AQ 1.9 m sorbent (Dr. Maisch HPLC

GmbH, Germany) in 360 m OD 100 m ID polyimide coated fused-silica capillary with a laser-

pulled emitter prepared on P2000 laser puller (Sutter, USA). Peptide separation was carried out

with a 2 h gradient of 80% ACN, 19.9% H2O, 0.1% FA (buffer B) in 99.9% H2O, 0.1% FA (buffer

A) at RT.

MS data were collected in DDA mode with the following parameters: MS1 resolution 70K,

3e6 AGC target with 30 msec IT, 350-2000 m/z scan range; MS2 resolution 17.5K, 1e5 AGC target

with 50 msec IT, 10 dependent MS2 scans, 1.4 m/z isolation window with 0.2 m/z offset, NCE 27;

min AGC 8e3, charge exclusion unassigned, 1 and >7, preferred peptide match with isotope

exclusion and 40 sec dynamic exclusion.

18

Raw files from BioPlex, from ProteomeExchange and the in-house generated LC-MS data

were subjected to protein identification in Peaks Studio X (Bioinformatic Solution Inc., Waterloo,

CA) against the UniProt SwissProt Homo Sapiens database containing both canonical and isoform

proteoforms (version from 2019.08.26) with a manually attached POLGARF sequence. Search

parameters included trypsin D|P digestion with max 3 miscleavages, precursor mass correction, 10

ppm and 0.05 Da error tolerance for precursor and fragment ions respectively, oxidation (M) and

deamidation (NQ) as variable modifications (max number of variable modification per peptide = 5)

and carbamidomethylation (C) as a fixed modification, Decoy-Fusion FDR estimation. Identification

results were filtered by 0.1% PSM FDR and 1 unique peptide per group with the final protein FDR

<1%.

Induction of POLGARFin secretion

In initial experiments Hek293T cells were transfected in three 4-well plates with POLGARF-

HiBiT (FuGene protocol). After every 24 hr a plate was taken out, medium and cells were harvested,

and luminescence in cell lysates and in the media was measured using Nano-Glo HiBiT luciferase

assay. This experiment was repeated 5 times.

To investigate the conditions that mediate POLGARFin secretion, Hek293T cells were

transfected with POLGARF-HibiT in 48-well plate (FuGene protocol). After 24 hr post transfection,

medium was substituted for a fresh DMEM with or without 10% FBS, with or without 100ug/ml of

cycloheximide (or the same volume DMSO as control). Every 30 min after media exchange,

aliquots of the media were collected and HiBiT activity was analysed with Nano-Glo HiBiT luciferase

assay (Promega). This experiment was repeated 21 times.

Purification of POLGARFin from cultured media and LC-MS analysis

To prepare the bait protein SNAP-LgBiT containing lysate, pcDNA3.4 SNAP-LgBiT plasmid

was transfected into expi293F cells (Expifectamine 293 protocol). 48 h after transfection, 25x106

cells expressing SNAP-LgBit were lysed in 5ml CCL buffer (Promega). The lysate aliquotes were

stored at -80oC.

To prepare the conditioned media for POLGARFin purification, expi293F cells were

transfected with pcDNA3.4 POLGARF-HiBiT (and pcDNA3.4 for negative control, Expifectamine

293 protocol). 48 h after transfection, the media was substituted for a fresh one, and transfected

cells were grown for additional 48h. Then the cells were sedimented by centrifugation and the

conditioned media was used for POLGARFin purification.

19

First, SNAP-LgBiT protein was loaded onto SNAP magnetic beads. For that, 40 l of

prewashed SNAP magnetic beads were incubated with 500 l of SNAP-LgBiT lysate for 1.5 h at

room temperature. Then the beads were washed in the CCL buffer supplemented with 0.5 M NaCl.

Next, SNAP-LgBiT-modified magnetic beads were incubated with 10 mL of the conditioned medium

for 30 min at room temperature with shaking. This procedure was repeated 5 times with fresh

aliquots of the POLGARF-HiBiT-transfected cells medium. After the final incubation, the magnetic

beads were washed in buffer A1000 (20 mM TrisHCl, 10% glycerol, 1mM DTT, EDTA, 1M KCl) and

then washed 2 times in PBS (we have found that high salt wash do not disrupt association of HiBiT-

containing proteins from SNAP-LgBiT). Bound proteins were eluted from the beads by incubation

with 1xSDS-Loading dye for 5 min at 50oC and resolved on Tris-Tricine SDS PAGE. The gels were

either stained with Coomassie (for mass spectrometry analysis) or transferred to nitrocellulose

membrane and then analysed with HiBit-blot. Briefly, the membrane was incubated in TBS-T buffer

for 30 min, followed by sequential incubation with the buffer, LgBiT protein, and Furimazine from

Nano-Glo HiBiT luciferase assay (Promega) for 5 minutes followed by chemiluminescence

detection on ChemiDoc XRS+ (BioRad).

For LC-MS sequencing, the HiBiT-containing protein was purified from the conditioned

media using the SNAP-LgBiT bait protein, resolved on a Tris-Tricine gel, and stained with

Coomassie. The protein band corresponding to the mobility of the POLGARFin-HiBiT protein in the

blot experiment of the same eluate, was excised and analysed by LC-MS/MS. In-gel protein

digestion was done as in (54) without protein reduction and Cys alkylation. Gel band was divided

into six pieces, three were digested with GluC and the other three with GluC and trypsin. All

samples were analysed by LC-MS in the same way as for the full-cell proteome analysis described

above. Peptides were separated on an Acclaim Pepmap C18, 2 um column 75 um *150 mm using

a 45 min gradient. Each sample was analyzed by a single LC-MS run.

All the obtained data files were analyzed together in PEAKS software against a Uniprot

human database with a manually attached POLGARF sequence and the common contaminant

database. Enzyme parameter was selected to be “Specified by each sample” with semispecific

digestion. Samples digested with GluC were analyzed as digested with “GluC (bicarbonate)”.

Samples digested with GluC and trypsin were analyzed with the digestion specificity combined from

“GluC (bicarbonate)” and “Trypsin”. The results were filtered to PSM FDR 1%, which resulted in

2.1% peptide FDR and 0.2 % protein group FDR. After FDR filtering, 6 runs resulted in identification

of 2617 PSM, 523 peptides and 53 protein groups, with most of the protein groups belonging to

common contaminants.

Riboseq data analysis

20

Publicly available Ribo-Seq/RNA-Seq datasets were downloaded from the NCBI Sequence

Read Archive and converted to FASTQ format, see Table S3. After adapter and rRNA removal,

reads below 25 nucleotides were discarded. Reads were then aligned to the GENCODE version

25 transcriptome using bowtie. Genomic positions of each read were determined and if aligning to

only one genomic position were classed as unambiguous. A metagene profile was created for each

Ribo-Seq read length from each dataset, centred around the annotated start codons. The distance

between the start codon and the highest peak upstream of the start codon plus 3 nucleotides was

used as an offset from the 5' end of each read to infer the A-site. A triplet periodicity score was

determined for each read length in each dataset by calculating the number of reads aligning to the

same frame as the annotated CDS's versus the number of reads in the -1/+1 frame. The triplet

periodicity score was then defined as number of reads in the highest peak divided by the number

of reads in the second highest peak, so that the read lengths with the strong periodicity would

receive a score close to 1, while the read lengths with weak periodicity would receive a score close

to 0.

Data Presentation

The number of biological replicates for each experiment is indicated in each figure legend.

For all Boxplots centre lines show the medians; box limits indicate the 25th and 75th percentiles as

determined by R software; whiskers extend 1.5 times the interquartile range from the 25th and 75th

percentiles, outliers are represented by dots. Genomic sequence alignments of 100 vertebrates

were visually explored using CodAlignView (I. Jungreis, M. Lin and M. Kellis, manuscript in

preparation) which was also used for generating Fig. 1C lower panels. The alignment shown in Fig.

3B was manually refined.

Data Availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange

Consortium via the PRIDE partner repository with the dataset identifier PXD016007

Acknowledgments

INS, DEA and MSM acknowledge Russian Science Foundation (RSF16-14-10065); PVB is funded

by SFI-HRB-Wellcome Trust Biomedical Research Partnership (Investigator Award in Science

(210692/Z/18/Z); MALDI MS facility became available to us in the framework of the Moscow State

University Development Program PNG 5.13; Funding for open access charge: Russian Science

Foundation

21

References

1. Kozak M (1980) Evaluation of the "scanning model" for initiation of protein synthesis in

eucaryotes. Cell 22(1 Pt 1):7-8.

2. Cigan AM, Feng L, & Donahue TF (1988) tRNAi(met) functions in directing the scanning

ribosome to the start site of translation. Science 242(4875):93-97.

3. Peabody DS (1989) Translation initiation at non-AUG triplets in mammalian cells. The

Journal of biological chemistry 264(9):5031-5035.

4. Kozak M (1999) Initiation of translation in prokaryotes and eukaryotes. Gene 234(2):187-

208.

5. Noderer WL, et al. (2014) Quantitative analysis of mammalian translation initiation sites by

FACS-seq. Molecular systems biology 10:748.

6. Ingolia NT, Ghaemmaghami S, Newman JR, & Weissman JS (2009) Genome-wide

analysis in vivo of translation with nucleotide resolution using ribosome profiling. Science

324(5924):218-223.

7. Ingolia NT, Lareau LF, & Weissman JS (2011) Ribosome profiling of mouse embryonic

stem cells reveals the complexity and dynamics of mammalian proteomes. Cell 147(4):789-

802.

8. Lee S, et al. (2012) Global mapping of translation initiation sites in mammalian cells at

single-nucleotide resolution. Proceedings of the National Academy of Sciences of the

United States of America 109(37):E2424-2432.

9. Fritsch C, et al. (2012) Genome-wide search for novel human uORFs and N-terminal

protein extensions using ribosomal footprinting. Genome research 22(11):2208-2218.

10. Ivanov IP, Firth AE, Michel AM, Atkins JF, & Baranov PV (2011) Identification of

evolutionarily conserved non-AUG-initiated N-terminal extensions in human coding

sequences. Nucleic acids research 39(10):4220-4234.

11. Van Damme P, Gawron D, Van Criekinge W, & Menschaert G (2014) N-terminal

proteomics and ribosome profiling provide a comprehensive view of the alternative

translation initiation landscape in mice and men. Molecular & cellular proteomics : MCP

13(5):1245-1261.

12. Tzani I, et al. (2016) Systematic analysis of the PTEN 5' leader identifies a major AUU

initiated proteoform. Open biology 6(5).

13. Hinnebusch AG, Ivanov IP, & Sonenberg N (2016) Translational control by 5'-untranslated

regions of eukaryotic mRNAs. Science 352(6292):1413-1416.

22

14. Johnstone TG, Bazzini AA, & Giraldez AJ (2016) Upstream ORFs are prevalent

translational repressors in vertebrates. The EMBO journal 35(7):706-723.

15. Calvo SE, Pagliarini DJ, & Mootha VK (2009) Upstream open reading frames cause

widespread reduction of protein expression and are polymorphic among humans.

Proceedings of the National Academy of Sciences of the United States of America

106(18):7507-7512.

16. Dever TE, et al. (1992) Phosphorylation of initiation factor 2 alpha by protein kinase GCN2

mediates gene-specific translational control of GCN4 in yeast. Cell 68(3):585-596.

17. Andreev DE, et al. (2015) Translation of 5' leaders is pervasive in genes resistant to eIF2

repression. eLife 4:e03971.

18. Vattem KM & Wek RC (2004) Reinitiation involving upstream ORFs regulates ATF4 mRNA

translation in mammalian cells. Proceedings of the National Academy of Sciences of the

United States of America 101(31):11269-11274.

19. Sidrauski C, McGeachy AM, Ingolia NT, & Walter P (2015) The small molecule ISRIB

reverses the effects of eIF2alpha phosphorylation on translation and stress granule

assembly. eLife 4.

20. Wiese A, Elzinga N, Wobbes B, & Smeekens S (2004) A conserved upstream open reading

frame mediates sucrose-induced repression of translation. The Plant cell 16(7):1717-1729.

21. Hill JR & Morris DR (1993) Cell-specific translational regulation of S-adenosylmethionine

decarboxylase mRNA. Dependence on translation and coding capacity of the cis-acting

upstream open reading frame. The Journal of biological chemistry 268(1):726-731.

22. Ivanov IP, et al. (2018) Polyamine Control of Translation Elongation Regulates Start Site

Selection on Antizyme Inhibitor mRNA via Ribosome Queuing. Molecular cell 70(2):254-

264 e256.

23. Trifunovic A, et al. (2004) Premature ageing in mice expressing defective mitochondrial

DNA polymerase. Nature 429(6990):417-423.

24. Michel AM, et al. (2014) GWIPS-viz: development of a ribo-seq genome browser. Nucleic

acids research 42(Database issue):D859-864.

25. Kolitz SE, Takacs JE, & Lorsch JR (2009) Kinetic and thermodynamic analysis of the role

of start codon/anticodon base pairing during eukaryotic translation initiation. Rna

15(1):138-152.

26. Diaz de Arce AJ, Noderer WL, & Wang CL (2018) Complete motif analysis of sequence

requirements for translation initiation at non-AUG start codons. Nucleic acids research

46(2):985-994.

23

27. Hinnebusch AG (2011) Molecular mechanism of scanning and start codon selection in

eukaryotes. Microbiology and molecular biology reviews : MMBR 75(3):434-467, first page

of table of contents.

28. Loughran G, Sachs MS, Atkins JF, & Ivanov IP (2012) Stringency of start codon selection

modulates autoregulation of translation initiation factor eIF5. Nucleic acids research

40(7):2898-2906.

29. Ivanov IP, Loughran G, Sachs MS, & Atkins JF (2010) Initiation context modulates

autoregulation of eukaryotic translation initiation factor 1 (eIF1). Proceedings of the

National Academy of Sciences of the United States of America 107(42):18056-18060.

30. Rosenbloom KR, et al. (2015) The UCSC Genome Browser database: 2015 update.

Nucleic acids research 43(Database issue):D670-681.

31. Lin MF, Jungreis I, & Kellis M (2011) PhyloCSF: a comparative genomics method to

distinguish protein coding and non-coding regions. Bioinformatics 27(13):i275-282.

32. Gao Y, Ma J, Saghatelian A, & Yates JRI (2017) Targeted searches for novel peptides in

big mass spectrometry data sets. BioRxiv.

33. Huttlin EL, et al. (2017) Architecture of the human interactome defines protein communities

and disease networks. Nature 545(7655):505-509.

34. Johnson M, et al. (2008) NCBI BLAST: a better web interface. Nucleic acids research

36(Web Server issue):W5-9.

35. Bekker-Jensen DB, et al. (2017) An Optimized Shotgun Strategy for the Rapid Generation

of Comprehensive Human Proteomes. Cell systems 4(6):587-599 e584.

36. Kulak NA, Geyer PE, & Mann M (2017) Loss-less Nano-fractionator for High Sensitivity,

High Coverage Proteomics. Molecular & cellular proteomics : MCP 16(4):694-705.

37. Rieckmann JC, et al. (2017) Social network architecture of human immune cells unveiled

by quantitative proteomics. Nature immunology 18(5):583-593.

38. Jiang J, Zhang Y, Krainer AR, & Xu RM (1999) Crystal structure of human p32, a doughnut-

shaped acidic mitochondrial matrix protein. Proceedings of the National Academy of

Sciences of the United States of America 96(7):3572-3577.

39. Fogal V, et al. (2010) Mitochondrial p32 protein is a critical regulator of tumor metabolism

via maintenance of oxidative phosphorylation. Molecular and cellular biology 30(6):1303-

1318.

40. Yagi M, et al. (2012) p32/gC1qR is indispensable for fetal development and mitochondrial

translation: importance of its RNA-binding ability. Nucleic acids research 40(19):9717-

9737.

41. Schwinn MK, et al. (2018) CRISPR-Mediated Tagging of Endogenous Proteins with a

Luminescent Peptide. ACS chemical biology 13(2):467-474.

24

42. Hinnebusch AG (2014) The scanning mechanism of eukaryotic translation initiation.

Annual review of biochemistry 83:779-812.

43. Schmitz J (2012) SINEs as driving forces in genome evolution. Genome dynamics 7:92-

107.

44. Hubley R, et al. (2016) The Dfam database of repetitive DNA families. Nucleic acids

research 44(D1):D81-89.

45. Wright PE & Dyson HJ (2015) Intrinsically disordered proteins in cellular signalling and

regulation. Nature reviews. Molecular cell biology 16(1):18-29.

46. Berlow RB, Dyson HJ, & Wright PE (2018) Expanding the Paradigm: Intrinsically

Disordered Proteins and Allosteric Regulation. Journal of molecular biology 430(16):2309-

2320.

47. Richards S, et al. (2015) Standards and guidelines for the interpretation of sequence

variants: a joint consensus recommendation of the American College of Medical Genetics

and Genomics and the Association for Molecular Pathology. Genetics in medicine : official

journal of the American College of Medical Genetics 17(5):405-424.

48. Grentzmann G, Ingram JA, Kelly PJ, Gesteland RF, & Atkins JF (1998) A dual-luciferase

reporter system for studying recoding signals. Rna 4(4):479-486.

49. Subach FV, et al. (2009) Monomeric fluorescent timers that change color from blue to red

report on cellular trafficking. Nature chemical biology 5(2):118-126.

50. Bensaddek D, Nicolas A, & Lamond AI (2016) Quantitative Proteomic Analysis of the

Human Nucleolus. Methods in molecular biology 1455:249-262.

51. Desiere F, et al. (2006) The PeptideAtlas project. Nucleic acids research 34(Database

issue):D655-658.

52. Wessel D & Flugge UI (1984) A method for the quantitative recovery of protein in dilute

solution in the presence of detergents and lipids. Analytical biochemistry 138(1):141-143.

53. Kovalchuk SI, Jensen ON, & Rogowska-Wrzesinska A (2019) FlashPack: Fast and Simple

Preparation of Ultrahigh-performance Capillary Columns for LC-MS. Molecular & cellular

proteomics : MCP 18(2):383-390.

54. Shevchenko A, Tomas H, Havlis J, Olsen JV, & Mann M (2006) In-gel digestion for mass

spectrometric characterization of proteins and proteomes. Nature protocols 1(6):2856-

2860.

55. Kozak M (1990) Downstream secondary structure facilitates recognition of initiator codons

by eukaryotic ribosomes. Proceedings of the National Academy of Sciences of the United

States of America 87(21):8301-8305.

56. Martinez HM (1990) Detecting pseudoknots and other local base-pairing structures in RNA

sequences. Methods in enzymology 183:306-317.

25

Figures and Tables

Figure 1. Identification of an alternative translated ORF in POLG

А. Schematic representation of reporter constructs bearing full length POLG 5’leaders (with and

without the uORF AUG codon), its AUG start codon plus 33 nt fused to firefly luciferase (in frame

with the POLG AUG), and relative luciferase activities of corresponding constructs transfected

into HEK293T cells. B. Riboseq aligned to each reading frame for the POLG mRNA generated in

the Trips-Viz browser. The region in the -1 frame which has high Riboseq density is depicted by a

green dotted box. C. GWIPS-Viz tracks for Riboseq (red) and RNAseq (green) global aggregates

for exon 3 of POLG (top). Lower panels represent CodAlignView alignment of 100 vertebrate

genomes (hg38/100) for the -1 frame (middle) and frame 0 (bottom).

26

Figure2. POLGARF translation from a CUG codon

A. Sequence logo of multiple alignments of a POLG 5’leader fragment generated for 85 placental

mammals. Sequence below represents nucleotide positions that are universally conserved. The

POLGARF CUG codon and surrounding nucleotides are highlighted in red. R and Y stands for

purines and pyrimidines, respectively. B. Schematic representation of reporter constructs tested

and results of transfection in HEK293T cells. The lower panel represents the results of Western

blotting against Firefly Luciferase and control Renilla luciferase (co-transfected with constructs

tested). C. Luciferase reporter assay of selected constructs co-transfected with plasmids

expressing EIF1 or EIF5. 1 represents constructs derived from construct 1 in B. 6 = construct 6

from B.

27

Figure 3. Conservation and evolution of POLGARF

A. PhyloCSF tracks for the POLG gene. POLGARF is indicated with a thick black line above

Refseq annotation. RepeatMasker tracks for POLG exons are below ((CTG)n corresponds to a

poly-glutamine track in POLG for minus strand). Reading frames for POLG and POLGARF ORFs

are indicated with blue and green stars respectively. B. Possible origin of POLGARF. A genomic

alignment is shown for representative mammalian species. Dfam/Repbase MirC consensus is

given below. Nucleotides that differ from MirC sequence are highlighted: common among

marsupials and placental mammals are in green; marsupial variants incompatible with POLGARF

expression are in red; the rest are in black. POLG CDS sequence is in yellow. The timing of MirC

sequence transposition is shown in the tree as a green cloud, the likely timing of mutations that

led to the highlighted variants (in green and red) are indicated in the tree with the same colours.

The initiation starts for POLG and POLGARF are indicated with rectangles.

28

Figure 4. POLGARF interaction, localization and functional analysis.

A. Coomassie blue stained SDS PAGE of a GST-POLGARF pull down assay. The major protein

bands were excised and subject to mass spectrometry analysis. B. SNAP pull down assay from

cells transfected with indicated constructs. Anti-FLAG and anti-POLGARF antibodies (see also

Fig. S9) were used for Western blotting analysis. C and D. Confocal imaging of cells transfected

with the indicated constructs and stained with corresponding SNAP & CLIP dyes and antibodies.

Line profile analysis across representative cells shows localization of fluorescent signals. In C,

yellow arrowheads show localization of RNA polymerase I (unstained spots), surrounded by

POLGARF. E. Coomassie blue stained SDS-PAGE with total cytoplasmic lysate and with

nucleolus-enriched fractions of cells transfected with the indicated constructs. Protein bands

corresponding to P32 and POLGARF were excised and subject to mass spectrometry.

29

Figure 5. POLGARF cleavage and secretion of its C-terminal fragment upon serum

stimulation.

A. HiBiT-blotting of cell lysate after expression of POLGARF-HiBiT. B. HiBiT-blotting of secreted

HiBiT containing polypeptide purified from conditioned media from cells expressing POLGARF-

HiBiT. C. LC-MS sequencing of secreted POLGARFin-HiBiT. Peptide peak areas are shown for

the sequences at the ragged N-term of POLGARFin. D. Measurement of HiBiT activity in the

conditioned media from POLGARF-HiBiT transfected Hek293T cells. 24h after transfection, the

media was replaced with fresh DMEM +/- 10% FBS and +/- cycloheximide, after indicated time

points the media aliquotes were assayed with Nano-Glo Nano-Glo HiBiT luciferase assay (n=24).